Impact of changes to the delivery of opioid agonist treatment, introduced during the COVID-19 pandemic, on treatment access and dropout in Ireland: An interrupted time series analysis

Background Following the emergence of COVID-19, Ireland introduced national contingency guidelines to ensure rapid and uninterrupted access to opioid agonist treatment (OAT). This study aims to assess the impact of changes introduced to the delivery of OAT on the number of people accessing treatment and treatment dropout. Methods The study conducted interrupted time series analyses, with separate segmented regression models (March 2019–February 2020) vs (April 2020–March 2021), for (A) total number of people accessing OAT, (B) the number initiating treatment, and (C) the number dropping out of treatment, using data from the National OAT treatment register. The study examined immediate (change in level or intercept: β2) and long-term impacts (change in slope; i.e., the difference between the slope before and after the intervention: β3). We performed total and stratified analyses by gender, age group (<40/≥40 years), and OAT drug (methadone or buprenorphine). Results A total of 10,251 people accessed OAT in Ireland in March 2019 (2 % buprenorphine, n = 178), increasing to 11,441 (4 % buprenorphine, n = 471) in March 2021. The study observed an immediate (β2 = 504.3, p < 0.001) and continued (β3 = 31.9, p < 0.001) increase of people accessing treatment following the introduction of the OAT contingency guidelines. In contrast, observed changes in level and slope were not significant for treatment initiation or dropout. The study did find, however, a modest reduction in dropout among those receiving buprenorphine (β3 = −0.6, p = 0.036). Conclusions Changes introduced to the delivery of OAT, under the COVID-19 contingency guidelines, are associated with increased access to OAT in Ireland, with no evidence of increase in treatment dropout. Whether these effects will be maintained over time remains to be seen.


Introduction
Opioid use disorder (OUD) is a major public health problem, contributing substantially to the global burden of disease, with an estimated 40.5 million people dependent on opioids globally (Degenhardt et al., 2019). The most recent estimates of opioid use in Ireland suggest a rate of 6.18 per 1000 population aged 15-64 years (Millar, 2018). Methadone is the most common form of opioid agonist treatment (OAT) in Ireland, which is available free of charge to all persons undergoing OAT for OUD. In 1998 the Misuse of Drugs (Supervision of Prescription and Supply of Methadone) Regulations were introduced in Ireland, which involved the establishment of a national treatment register, the Central Treatment List (CTL). The 1998 Regulations were updated in 2017 to provide for OAT with buprenorphine. All patients in receipt of OAT, including methadone, are listed on the CTL, with each patient linked to one specific prescriber and a single dispensing site. OAT is provided in specialist outpatient addiction clinics or primary care settings, with approximately 60 % of people in treatment in addiction clinics (Delargy et al., 2019;Durand et al., 2021).
The emergence of COVID-19 presented significant challenges to the provision of OAT within existing regulations and clinical guidelines, as OAT is heavily dependent on regular face-to-face health care delivery (Marsden et al., 2020). At the beginning of the pandemic, real concerns existed that disruption to care, particularly access to OAT and other prescribed medication, would have detrimental consequences for people undergoing treatment (Volkow, 2020). Furthermore, research anticipated that many people would seek treatment during COVID-19 due to disruptions to the supply of illicit opioids (Marsden et al., 2020). This led to a rapid and coordinated response to mitigate the spread of COVID-19, while ensuring continued and safe access to OAT across multiple sectors of the Irish Health and Social Care system, which resulted in the introduction of a suite of national contingency guidelines in March 2020. These contingency guidelines supported accelerated access to OAT for people not already in treatment, increased access to buprenorphine, and increased take-home doses where clinically appropriate and harm reduction measures such as naloxone prescribing. E-consultations, electronic prescriptions, and home delivery of medications for those selfisolating were also set up to facilitate continuity of OAT (Hennigan et al., 2021).
Rapid access to OAT is an important marker of quality of patient care, and COVID-19 may have created an opportunity to increase access to OAT in Ireland. However, growing evidence suggests that the risk of mortality following dropout from OAT is high (Cousins et al., 2016;Durand et al., 2020;Santo et al., 2021); therefore, research must examine treatment dropout, alongside the number initiating treatment during the pandemic. The aim of this study is to evaluate the impact of changes introduced to the delivery of OAT on the total number of people receiving OAT, the number initiating OAT, and the number dropping out of treatment, using an interrupted time series (ITS) design.

Study design and setting
We used an ITS study design, with anonymized aggregated level data from the national register of people receiving OAT in Ireland, the CTL. People are registered on the CTL when they start OAT, and their treatment status remains active on the CTL for up to 28 days from their first day of nonattendance with their treatment provider. During this time, attempts are made to contact the person to encourage them to return to treatment. If no contact is made and/or the person does not attend for 28 consecutive days, they are recorded as having exited treatment. As a mandatory national register, the aggregated number from the CTL is nationally representative. The RCSI Research Ethics Committee approved this study (REC202009008). This study is reported according to the STROBE Standardized Reporting Guidelines for Cross-Sectional Studies (Vandenbroucke et al., 2007). The study protocol is available on HRB Open .

Primary outcomes
The primary outcomes of interest were: (A) total number of people accessing OAT on the last day of each month; (B) number of people initiating OAT each month (including first-timers and those re-entering treatment following a period of >28 days off treatment); and (C) number of people dropping out of OAT each month, identified as having exited the CTL. Dropout excludes deaths and transfers to other settings, such as hospitals or prison. The study observed rates of dropout, accounting for the number of people in treatment on the last day of the previous month, to detect any variation to outcome (C). While we originally planned to assess the impact of changes introduced to the delivery of OAT on number of people on waiting lists and waiting time , the required data were not available.

Statistical analysis
We used monthly aggregated CTL data from March 2019 to March 2021. The study team selected this time frame to include a sufficient runin phase before the changes were introduced (pre-intervention: March 2019 to February 2020), as well as a 12-month follow-up phase (postintervention: April 2020 to March 2021). March 2020 was removed from the analysis as the contingency guidelines were introduced midmonth.
We conducted separate segmented regression models for each outcome: Y t represents the dependent variable; β 0 the baseline level precontingency guidelines; β 1 the precontingency guidelines trend; β 2 the change in level immediately following the contingency guidelines; β 3 the change in slope following the contingency guidelines (i.e., the deviation from the existing trend), and e t represents the unexplained variability. The study used Newey-West robust standard error (Newey & West, 1987) with lag(0) a priori in the models. Autocorrelation and partial autocorrelation function plots were visually inspected, and we used the Cumby-Huizinga test (Cumby & Huizinga, 1992) to identify the presence of autocorrelation in the regression model residuals. Autocorrelation is a common phenomenon in time series data in which there is a correlation between the value of a variable at a given time, and the value of this same variable at an earlier time. In the presence of lag-1 autocorrelation (i.e., correlation between the values on month m and m + 1) in the model residuals, the study applied a generalized least-squares transformation (Prais-Winsten) (Beard et al., 2019;Lopez Bernal et al., 2016;Prais & Winsten, 1954). We assumed a significance level α = 0.05.
Study staff also performed subgroup analyses by gender, age (<40 years/≥40 years) and OAT drug (methadone or buprenorphine). We identified outliers using a Grubbs test (Grubbs, 1969), and where appropriate we carried out sensitivity analysis removing outliers. We used Stata/SE v16.0 for data analysis.
Frequencies and percentages are reported for the total number of people accessing OAT at the beginning and end of the study period (March 2019 and March 2021). Median and inter-quartile range (IQR) are reported for outcomes (B) and (C). Model regression coefficients (β 0, β 1 , β 2 and β 3 ), associated 95 % confidence interval (95 % CI) and significance values are presented for each outcome. β 2 and β 3 coefficients different from zero indicate an immediate (change in level) and continued (change in slope) change in the outcome measured.
As Table 1 shows, the study found a significant monthly increase in number of people accessing OAT prior to the contingency guidelines. Table 1 and Fig. 1A, also demonstrate a significant positive change in level (β 2 = +504.3, 95 % CI 441.6-567.1, p < 0.001) and slope (β 3 = +31.9, 95 % CI 24.5-39.4, p < 0.001), reflecting an initial increment of approximately 500 people at the time of the intervention and an acceleration in the growth of the population in treatment in the following months.
Similar results were observed across gender and for those prescribed buprenorphine. The study observed a negative change in slope for methadone (β3 = − 12.0, 95 % CI -23.8 to − 0.1, p = 0.048). However, it did not remain significant when we removed April 2020 (identified as an outlier). A negative change in slope was also observed for those aged ≥40 years (β3 = − 3.9, 95 % CI -6.9 to − 0.8, p = 0.017), reflecting a decrease in the number of people ≥40 initiating treatment in the months following the intervention (Supplementary Table 2).

Number of people dropping out of OAT
The median (IQR) monthly dropout number was 69.5 (66.5-79.5) over the study period. Although the study observed an increase in the number dropping out of OAT prior to the contingency guidelines (β 1 = +1.1, 95 % CI 0.1-2.1, p = 0.036), it observed no significant changes in level or slope (Table 1). Similar effects were found across gender, age and those prescribed methadone. The study found a significant, negative change in slope for those prescribed buprenorphine (β3 = − 0.6, 95 % CI -1.2 to − 0.1, p = 0.036, Supplementary Table 3), reflecting a deviation (reduction) from the pre-existing trend for the number of people dropping out after the intervention. When we considered rates of dropout, accounting for the number of people in treatment each month, the results remained largely unchanged. The only observed difference was the rate of dropout in the buprenorphine group, prior to the contingency guidelines, for which the increase did not remain significant.

Discussion
Our findings would suggest that changes introduced to the delivery of OAT facilitated increased coverage of OAT in Ireland, an important protective factor for reducing overdose and all-cause mortality (Chaillon et al., 2022;Santo et al., 2021). While the study found a peak in the number of people starting treatment in April 2020, 59 % of whom had prior treatment experience, this increase in starting or re-entering OAT was not sustained in the short or long term. One explanation for this finding is that the pool of people who were on the waiting list was cleared, which reduced demand in the following months. The immediate and sustained increase in OAT coverage in Ireland arises from the initial increase in people entering treatment and the decrease or absence of change in the number of people dropping out of OAT, particularly among those on buprenorphine. The literature suggests similar to lower retention rates on buprenorphine vs methadone overall (Mattick et al., 2014). Therefore, it would have been important to monitor dropout levels beyond our study period. Also of note, dropout was lowest in April 2020 and January 2021, coinciding with "hard" lockdown measures in Ireland (closure of all nonessential retail, restriction of movement, police enforcement). This finding is consistent with recent qualitative studies, which suggest that people's motivation to initiate and remain in treatment during the pandemic was influenced, in part, by supply interruptions for illicit drugs, and an inability to secure income to purchase drugs (May et al., 2022). Therefore, as COVID-19 restrictions continue to ease, and life returns to normal, dropout rates may begin to increase, leading to increased risk of overdose deaths, undermining the gains achieved in improving treatment access. As previously suggested, a clinical decision support system, identifying peoples' risk of dropout in real time, could support targeted service enhancements to increase patient engagement and prevent dropout (Santo et al., 2021).
The contingency guidelines for OAT recommended multiple changes to the delivery of OAT. Which specific changes, or combination of changes, led to the observed effects is unclear. A recent Canadian study suggests that increased take-away doses of OAT was associated with lower rates of treatment interruption and discontinuation at six-months, and was not associated with an increase in overdose deaths during the same time period (Gomes et al., 2022). In addition, we selected March 2020 as the "intervention period". Some services may have changed their practices prior to the publication of these guidelines, having observed the situation in other countries, such as in China and Italy. We used the total number of people on treatment on the last day of the month to represent the total number of people on OAT each month. Although a potential underestimation, we applied this consistently throughout the observation period and it closely reflects the dynamic of the total number of people attending OAT each month. Furthermore, while the CTL is nationally representative of people in OAT in Ireland, it is a dynamic database that is constantly updated to account for people leaving the register due to death, departing the country, or dropping out of treatment. Delays in updating exit details for individuals may lead to information bias, particularly toward the end of the study period. In effect, such bias would likely lead to either underestimates (delayed reporting) or overestimates (misclassification of deaths as dropout) of the number of dropouts. However, the database is validated at the end of each month, and we expect limited impact.
Waiting time before treatment initiation is an important indicator for  quality of care and should be reported when data become available. In addition, individual data would be useful to capture personal experience and identify factors associated with dropout, e.g., for the people driven by the pandemic circumstances to access OAT for the first time. Last, due to the nature of death investigations and the impact of COVID-19 on data collection processes from Coronial files (Lynn et al., 2020(Lynn et al., , 2021, data on overdose mortality are not currently available for the studied years. Once these data become available, research should assess the impact of OAT delivery changes on mortality, particularly drug overdose deaths.

Conclusion
Our findings suggest that changes introduced to the delivery of OAT, documented in the contingency guidelines, resulted in increased access to OAT in Ireland, and, in the case of buprenorphine, a sustained reduction in treatment dropout. Whether these effects will be maintained over time and what impact this had on mortality, including overdose, remains to be seen.

Role of funding source
This study was funded by the Health Research Board under the Research Collaborative in Quality and Patient Safety (RCQPS), a collaborative initiative between the Health Research Board, the Health Service Executive, National Quality Improvement Team and the Royal College of Physicians of Ireland [RCQPS-2020-016]. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

CRediT authorship contribution statement
LD extracted and managed the data, performed the analyses interpreted the results and drafted the manuscript. FB supported the data analyses revised the manuscript and provided critical comments. NH, ID, MB & MOV revised the manuscript and provided critical comments. WE facilitated access to the data, revised the manuscript and provided critical comments. EK interpreted the results, revised the manuscript and provided critical comments. GC (PI) designed the study, secured the data, interpreted the results and drafted the manuscript. All authors approved of the final manuscript.

Declaration of competing interest
No conflict declared.